Visual Mapping of Internal Pipe Walls using Sparse Features for Application on board Autonomous Underwater Vehicles

In this project an algorithm to generate a single image of an entire water pipe's inside wall was developed. An Autonomous Underwater Vehicle (AUV) equipped with a fish-eye camera will be deployed to dive through the pipe and take pictures of its inside wall. The algorithm was implemented that maps such pictures to the three dimensional model of the water pipe and based on that, generates the image that one would see if the photographed section of pipe was cut along the side and rolled out. The algorithm then uses the rolled out versions of a series of consecutively taken photos to generate a mosaic showing the entire pipe in a single picture based on the recognition of sparse, structural features. The performance of the algorithm was verified in land based experiments and the robustness of the system to measurement inaccuracy was assessed. Finally two mosaicked images were created for different paths taken through the pipe, to demonstrate that the system is capable of generating the same image, regardless of the path taken by the AUV.